Can You Grade EL Defects by Grayscale? What It Takes for Brightness to Become a Reliable Severity Scale

ⓘ This article is currently published in Chinese. An English version is being prepared — read the Chinese original.

A growing number of EL inspection devices and software now advertise automatic defect grading 'by grayscale value'—reading how dark a region is and mapping that darkness onto a severity grade. The idea is appealing because it promises an objective number instead of a human judgment call. But grayscale in an EL image is not an absolute property of the module; it is the end result of a chain of settings—injection current, exposure time, camera gain and dark noise, lens vignetting, shooting distance and angle. Change any of them and the same crack can read brighter or darker. This article explains why raw grayscale is not directly comparable across devices or across sessions, what normalization it takes before brightness can serve as a quantitative severity scale (fixed and documented imaging parameters, in-frame healthy-cell references, calibration against known samples, metrological traceability), and where grayscale grading legitimately helps versus where a dark region is simply by design rather than a defect. The conclusion: grayscale can become a measurable severity reference, but only on top of fixed parameters, normalization, an agreed defect-grading rule, and traceability—otherwise 'quantitative grading' is just a numeric coat thrown over subjective reading. AI auto-grading does not remove this prerequisite; it inherits it.